Hindawi Publishing Corporation Journal of Electrical and Computer Engineering Volume 2013, Article ID 489515, 21 pages http://dx.doi.org/10.1155/2013/489515

Research Article A Systemic Approach to the Preservation of Audio Documents: Methodology and Software Tools

Federica Bressan1 and Sergio Canazza2

1 Department of Computer Science, University of Verona, Strada Le Grazie 15, 34134 Verona, Italy 2 Department of Information Engineering, University of Padova, via G. Gradenigo 6/B, 35131 Padova, Italy

Correspondence should be addressed to Federica Bressan; [email protected]

Received 10 December 2012; Revised 19 February 2013; Accepted 21 February 2013

Academic Editor: Filippo Stanco

Copyright © 2013 F. Bressan and S. Canazza. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

This paper presents a methodology for the preservation of audio documents, the operational protocol that acts as the methodology, and an original open source software system that supports and automatizes several tasks along the process. The methodology is presented in the light of the ethical debate that has been challenging the international archival community for the last thirty years. The operational protocol reflects the methodological principles adopted by the authors, and its effectiveness is based on the results obtained in recent research projects involving some of the finest audio archives in Europe. Some recommendations are given for the rerecording process, aimed at minimizing the information loss and at quantifying the unintentional alterations introduced by the technical equipment. Finally, the paper introduces an original software system that guides and supports the preservation staff along the process, reducing the processing timing, automatizing tasks, minimizing errors, and using information hiding strategies to ease the cognitive load. Currently the software system is in use in several international archives.

1. Introduction In an official technical report, the UNESCO [2]claims that over half of the cultural patrimony of the world is at seri- Computer science offers multiple possibilities to study the ous risk of vanishing, despite the attention that a topic such fields of humanities: a major topic that has been rapidly grow- as cultural heritage preservation has attracted, especially by ing along the past decades is the implementation of computer theEuropeanUnionwhichprovedgreatawarenessinfinanc- engineering in musical cultural heritage, with a particular ing a number of research projects in this field. relation to the audio documents preservation (“ideally, an The factors that obstruct the safeguard of audiovisual audio preservation workflow would also involve the services documents are multiple: mainly the massive investment of of a specialized programmer” [1]). human and economical resources required by digitization Scholars and the general public started paying greater campaigns, not to mention teams with multidisciplinary attention to the recordings of musical events and to their competences, difficult and expensive to form. As a conse- value, at a personal/collective level and for cultural/entertain- quence, today many archives are in fact lacking method- ment purposes. However, a systematic preservation and the ological and technological tools to safeguard adequately their fruition of these documents is complicated by their diversi- patrimony. fied nature: recordings contain information on their artistic The process of physical degradation that characterizes and cultural existence that goes beyond the audio signal itself. every type of audio carriers can be slowed down, by means In this sense, a faithful and satisfying access to the audio doc- of correct preservation policies, but not stopped. Therefore, ument cannot be achieved without its associated contextual the survival of the information contained in the document is information, that is, to all the content-independent informa- possible only renouncing to its materiality, through a constant tion represented by the container, the signs on the carrier, the transfer of the information onto new carriers. Unfortu- accompanying material, and so on. nately the remediation process (the operation of transferring 2 Journal of Electrical and Computer Engineering the information from a medium to another medium)issubject audio information stored in obsolete formats and carriers is in to electronic, procedural, and operative errors. In addition, risk of disappearing. To this end, the audio preservation com- it easily indulges to the aesthetic changes of the current munityintroducedtheconcept“preservethecontent,notthe times. Consequently, a total neutrality in the process of the carrier.” Audio (and video) preservation must therefore be information transfer is not realistic, which puts the spotlight based on digital copying of contents. Consequently, analogue on the philological problem of the documents authenticity. holdings must be digitized. At the end of the 20th century, The recording of an acoustic event can never be a neutral the traditional “preserve the original” paradigm shifted to operation because the timbre quality and the plastic value the “distribution is preservation” [4] idea of digitizing the of the recorded sound, which are of great importance in, for audio content and making it available using digital libraries example, contemporary music, are already influenced by the technology. Now the importance of transferring into the positioning of the microphones used during the recording. digital domain (active preservation) is clear, namely, for In addition, the processing carried out by the Tonmeister is carriers in risk of disappearing, respecting the indications of a real interpretative element added to the recording of the theinternationalarchivecommunity[8–12]. event. The term Tonmeister3 [ ] describes a person who has a For the longest time, the word “archive” has been suggest- detailed theoretical and practical knowledge of all aspects of ingimagesofshelvesandbooks.Itisonlyrecentlythataudio- soundrecording.But,unlikeasoundengineer,he/shemust visual material, first labeled as nonbook material (the only way be also deeply musically trained: musicological and historic- to define it was in contrast with books), earned the right to be critical competence are essential for the individuation and stored in a controlled environment, next to the books, and to correct cataloguing of the information contained in audio be regarded as worthy of permanent preservation. However, documents. Thus, musicological and historic-critical compe- their inclusion has not been completely painless. It forced the tence becomes essential for the individuation and correct cat- archives to update their storage facilities and to plan radically aloguing of the information contained in audio documents. new access strategies. Multimedia documents encompass a Beingmadeofunstablebasematerials,soundcarriersare large variety of carriers and formats and require special moresubjecttodamagecausedbyinadequatehandling.The equipment to access their content. Whereas traditional media commingling of a technical and scientific formation with and tools remained in the possession of artists and curators, historic-philological knowledge (important for the individ- the multimedia technical and technological load has defi- uation and correct cataloguing of the information contained nitely reduced the autonomy of these cultural actors [13]. in audio documents) becomes essential for preservative re- Traditional criteria for conservation—a document’s origi- recording operations, going beyond mere analog-to-digital nality, longevity, and inherent economic value—are not appli- (A/D) transfer: “the development of successful preservation cable to multimedia documents. Archivists were faced with strategies will require the cooperation of computer scientists, unprecedented problems that transcended their field of data storage experts, data distribution experts, fieldworkers, expertise, aggravated by the complex nature of multimedia librarians, and folklorists” [4]. documents and by their content, which cannot be accessed As far as audio memories are concerned, preservation is without mediation technology. divided in passive (also referred to as preventative preserva- In the last thirty years, the awareness for the preservation tion) and active preservation, which involves data transfer of audio documents has increased, allowing them in the onto new media. Passive preservation is further divided definition of cultural heritage. At the same time, the debate into indirect (covering the maintenance of proper storage on the ethics of preservation got more lively and the tools to conditions, establishing professional handling procedures, activate the preservation practices have become richer. More etc.) and direct (the carrier is treated in order to stabilize its recently, audio recordings have been recognized as an impor- physical condition, but its structure and composition are not tant documentary source for scientific research in the fields altered). of linguistics, history, sociology, musicology, and other dis- It is worth noting that, in the Eighties/Nineties of the 20th ciplines. Their dignity has been equaled to that traditionally century, expert associations were still concerned about the reserved to bibliographic sources, revealing once more the use of technology and digital storage media centrality of philological problems such as the authenticity for long-term preservation [5, 6]. They recommended re- and the authority of the documents. recording of endangered materials on analogue magnetic Approaching the archival reality from a viewpoint of the tapes,becauseof(a)rapidchangeandimprovementofthe methodological research and of the scientific reflection, the technology, and thus rapid obsolescence of hardware, digital authors have identified some critical situations in the man- format and storage media; (b) lack of consensus regarding agement of the processes involved in the preservation of the samplerate,bitdepth,andrecordformatforsoundarchiving; audio patrimony. They have tried to propose effective solu- (c) questionable stability and durability of the storage media. tions using original software tools, the design of which The digitization was considered primarily a method of pro- required (i) a reflection on the relevant characteristics of viding access to rare, endangered, or distant materials—not a each type of carrier and (ii) the subsequent definition of a permanent solution for preservation. Smith, still in 1999, sug- metadata set that meets the requirements of the methodology gested that digitization should be considered a means for proposed by the authors in Section 3 as well as the require- access, not preservation—“at least not yet” [7]. It is well ments expressed by the archives. In order to do so, the known that preserving the carriers and maintaining the dedi- authors have considered the documentation produced by cated equipment for their reproduction are hopeless. The previous research projects [14–16] and by some of the most Journal of Electrical and Computer Engineering 3 authoritative archives in Europe [17]. All of the tools describ- equipment used for the digitization, focusing on the high- ed in this paper are open source and accessible in the “soft- quality/high-cost/low-throughput cases. The authors believe ware” section of [18]. Promoting the circulation of these tools that the increased dimensionality of the data contained is a priority in the authors agenda because it implies the cir- within an audio digital library should be dealt with by means culation of the methodological principles that lie behind of automatic annotations. them; this is especially addressed to medium and small archives that belong to that category of cultural institutions suffering from scarcity of fundings, of personnel, and of tech- 2. The Preservation of Audio Documents nical/technological competences for preservation. From the viewpoint of the scientific research, these This section presents the most significant positions of the tools fall within the scope of the area of cultural interfaces. thirty-year-long debate that was built around the preserva- The element of innovation consists in the adoption of a tion of audio documents. The ethics of preservation is dis- systemicapproachtothedevelopmentofsoftwaretools cussed from the viewpoint of the multiple motivations for for quality control in the management of semiautomatized going digital, which result in different choices on the oper- procedures for the creation and the description of digital ational level (see Section 2.4). archivesforpreservation,forthesharingoftheinforma- tion, and for the maintenance of their integrity over time. 2.1. “Two Legitimate Directions”. The journal article that In other words, for the automatic control of concurrent started it all in 1980 bore the signature of William Storm, at processes. that time Assistant Director of the Thomas A. Edison Re-re- In fact, the great majority of the operations that charac- cording Laboratory Syracuse University Libraries [19]. The terize the archival routines are highly repetitive, and those article was posing the problem of standardizing the proce- involved in the active preservation of audio documents are dures for audio restoration for the very first time, and it no exception. As a consequence, the amount of time spent for became famous for the number of controversies it stimulated. the processing and the management of electronic files is con- Storm individuated two legitimate directions, two types siderable, and the well-known pathologies of attention related of re-recording which are suitable from the archival point of to repetitive activities and/or low level tasks may induce the view: (1) the sound preservation of audio history and (2) the human operator to introduce errors which have a cascade sound preservation of an artist. effect on the workflow, causing the malfunctioning ofthe The first type of re-recording (Type I) represents a level algorithms that check the internal consistence of the archive of reproduction defined “as the perpetuation of the sound of and the algorithms for information retrieval. Also for the an original recording as it was initially reproduced and heard long-term maintenance of the archive adequate tools are by the people of the era.” [19]. Storm’s contribution aimed at required, as on the one hand the periodical control of every shifting the archivist’s interest from the simple collecting of single documents is not sustainable and on the other hand a audio carriers to the information contained in the recording, sample check is not satisfactory. and at highlighting the double documentary value of re- Inparticular,thetoolsdevelopedbytheauthors(a) recording by proposing an audio-history sound preservation: reduce the duration of restoration/remediation sessions, (b) on the one hand, he wanted to offer a historically faithful dramatically reduce the time necessary to create the access reproduction of the original audio recording by extracting copies, and (c) they introduce a set of redundant automatized the sound content according to the historical conditions and controls that ensure data integrity. PSKit PreservationPanel technology of the era in which it was produced; on the other has been used for the preservation of audio documents in hand,hewantedtodocumentthequalityofsoundreception several projects listed in Section 5. offered by the recording and reproducing systems of the time. This work is the result of the experience matured in sev- These two instances, conceptually joined in a single type eral research/applied projects carried out by the authors on of re-recording, had induced Storm to prescribe the use of Digital Audio Archives and Audio Access (see Section 5). The original replay equipment. The aim of history preservation archives involved in these projects, some of which are mono- “is to first hear how records originally sounded to the general thematic (e.g., classical opera, electroacoustic music) and public.” some highly heterogeneous (e.g., speech recordings, ethno- The second type of re-recording (Type II) was presented musicological documents), represent significant examples by Storm as a further stage of audio restoration, as a more among which most categories of sound archives fall. ambitious research objective, conceived as a coherent devel- After a detailed overview on how this debate evolved since opment of Type I: “The knowledge acquired through audio- the Seventies inside the archival community on the preserva- history preservation provides the sound engineer with a tion of audio documents (Section 2), the paper describes the logical place to begin the next step—the search for the operational protocol defined, the processes undertaken, and true sound of an artist.” Type II is then characterized by the results obtained in several international audio documents the use of “playback equipment other than that originally preservation projects. In particular, in Section 2.6 and in intended so long as the researcher proves that the process Section 3 some guidelines are given, including recommen- is objective, valid, and verifiable” [19], with the intent of dations to the analogue-to-digital process finalized at the obtaining “the live sound of original performers,” transcend- minimization of the information loss and at automatically ing the limits of a historically faithful reproduction of the measuring the unintentional alterations introduced by the recording. 4 Journal of Electrical and Computer Engineering

2.2. “To Save History, Not Rewrite It”. The Guide [12]com- compensations of type B are commonly accepted and must— missioned by UNESCO reports the philosophical approach as Schuller¨ writes—be carried out, in type C they have to do save history, not rewrite it. The audio section is clearly influ- withtheareaofequalizations“usedtocompensatefornon- enced by the new formulations made by Schuller.¨ Schuller’s¨ linear frequency response, caused by imperfect historical works [11] move from a different methodological point of recording equipment and to eliminate rumble, needle noise, view, “which is to analyse what the original carrier represents, or tape hiss” [11]. These are operations which elude standard technically and artistically, and to start from that analysis in operational criteria and must therefore be rigorously docu- defining what the various aims of re-recording may be” [12]. mented by the restorer, who must write out accurate reports Regarding the reconstruction of the history of music percep- in which he specifies both the equipment and systems used as tion Schuller¨ states: “The only case where the use of original well as all the restoration phases. equipmentisjustifiedisintheexoticaimtoreconstructthe sound of a historical recording as it was heard originally.” 2.3. “Secondary Information:” The History of the Audio Docu- Insteadhepointsdirectlytowardsdefiningaprocedurewhich ment Transmission. The studies of George Brock-Nannestad guarantees the re-recording of the signal’s best quality by [20] are in line with the modeling of the degradations through limiting the audio processing to the minimum. Having set reverse engineering. In these studies he focused on the A/D aside the general philosophical themes, Schuller¨ goes on to an conversion of acoustic recordings (thus recordings made accurate investigation of signal alterations which he classifies before 1925) and, in particular, the strong line spectrum in two categories: intentional and unintentional. The former in the recording transfer function and unknown recording include recording, equalization, and noise reduction systems, speed. Brock-Nannestad goes back to the first studies in the while the latter are further divided into two groups: the ones acoustics of sound reproduction and to the scientific works caused by the imperfection of the recording technique of the of Miller [21],whomwemustrecallasthefirsttoattemptto time, resulting in various distortions and the ones caused retrieve the true sound once it had been recorded. In order to by misalignment of the recording equipment, for example, be consistent and have scientific value, the re-recording work wrong speed, deviation from the vertical cutting angle in requires a complete integration between the historical-critical cylinders, or misalignment of the recording in knowledge which is external to the signal and the objective [12]. knowledge which can be inferred by examining the carrier The choice whether or not to compensate for these alter- and the degradations highlighted by the analysis of the signal. ations reveals different re-recording strategies: “historical faithfulness can refer to various levels: Type A the recording as it was heard in its time, which is equivalent to Storm’sType I 2.4. Main Reasons for Digitization. As has been pointed out presented in the previous section; Type B the recording as in the Introduction, the fact that digitization might be the it has been produced, precisely equalized for intentional re- current best solution to the problem of carriers degradation cording equalizations, compensated for eventual errors has been gradually accepted by the international community. caused by misaligned recording equipment and replayed on To date it is widely understood, yet the reasons for “going modern equipment to minimize replay distortions” [12]. digital” can differ depending on the archive characteristics, Type B re-recording defines a historically faithful level of size, and policies. In this sections, the main reasons accounted reproduction that, from a strictly preservative point of view, in [22]aresummarized. is preliminary to any further possible processing of the signal. The first reason for digitization can be enhanced access These compensations use knowledge which is external tothe to materials that were previously unavailable or only par- audio signal; therefore, even in the operations provided for tially/locally available. Digitization, in this sense, is part of by Type B, there is a certain margin of interpretation because the democratic considerations promoted by archives, making a historical acquaintance with the document is called into public records more widely accessible. Access can be permit- question alongside with technical-scientific knowledge. For ted to the metadata and/or to the data, enabling/extending instance, to individuate the equalization curves of magnetic the availability to support educational and outreach projects, tapes or to determine the rotation speed of a record. Most of or a defined stock of research material. Digitization may also the information provided by Type B is retrievable from be aimed at implementing the “virtual re-unification” of col- thehistoryofaudiotechnology,whileotherinformationis lections and holdings from a single original location or instead experimentally inferable with a certain degree of pre- creator now widely scattered, or at creating a single point of cision. The re-recording work can thus be carried out with access to documentation from different institutions concern- a good degree of objectivity and represents an optimal level ing a special subject. within which the standard for a preservation copy can be A second reason for “going digital” is to promote and to defined. facilitate the access. The main purpose is to enable the use of After having established an operational criterion for pre- material (original manuscripts and archives, maps, museum servative re-recordings, based on stable procedures and artifacts, rare books, etc.) that cannot be consulted in its derived from an objective knowledge of the degradations, original form other than by visiting its specific repository, Schuller¨ individuated a third level of historically faithful because it has been damaged or because it is easier and more reproduction, type C: “The recording as produced, but with productive to access without computer enhancement tools additional compensation for recording imperfections caused like OCR (Optical Character Recognition) or text encoding bytherecordingtechniqueofthetime”[11]. While the for converted texts. Journal of Electrical and Computer Engineering 5

Preservation, on which this paper is focused, appears as performance—except for the carriers with a very poor start- the third possible reason for starting a digitization campaign. ing condition: these might not endure one/multiple playback However,theauthorsbelievethatitshouldbeconsidereda sessionsandbenolongerreadableaftertheprocess(forthese, preliminary step to the realization of the other objectives. The the optimization of the carrier before playback is crucial since main reason is that the documents that are accessed by the there is only one chance to extract the best signal). In general, final users are the result of a processing chain that usually source documents should be kept for future comparisons and starts from a digital preservation master, aligned with the for other purposes that depend on the evolution of technol- highest standards, and aware of the philological questions of ogy and that cannot be predicted to date. “Discarding an ori- accuracy and authenticity. ginal, no matter how many copies have been made, should never be undertaken lightly” [2,page13].Ifeachstepiscar- 2.5. A Proposal for a Shared Operational Protocol. Acentral ried out according to the protocol, the preservation copy will concept in the operational protocol here proposed is the fulfill the requirements of accuracy, reliability, and philolog- “preservation copy” of an audio document: it consists in ical authenticity[23]. Authenticity is, in fact, the result of a an organized data set that groups all the information rep- process, it cannot be evaluated by means of a boolean flag, and resented by the source document, stored, and maintained it is never limited to the document itself but extended to the as the preservation master (see Section 3.5 for details). The information/document/record system [24]. remediation process depicted in Figure 1 is composed by the totality of the steps from the evaluation of the state of pre- 2.6. Passive Preservation. Direct passive preservation can be servation of the source document to the completion of a pre- carried out only if the main causes of the physical carriers servation copy. It is divided in three main blocks (before play- deterioration are known and consequently avoided. We sum- back, playback, and after playback), each of which is articu- marize the main risks for the most common categories of car- lated in procedures and subprocedures. The output of each riers: mechanical carriers, magnetic tapes, optical discs, and procedure or sub-procedure is either data, a report, or a dif- magneto-optical carriers. ferentstateofthesystem.Thecompletesequenceisasfollows. (1) Preparation of the carrier 2.6.1. Mechanical Carriers. The common factor with this (1.1) Physical documentation group of documents is the method of recording the infor- mation,whichisobtainedbymeansofagroovecutintothe (1.1.1) Photographic documentation surface by a stylus modulated by the sound, either directly (1.1.2) Scanned images in the case of acoustic recordings or by electronic amplifiers. (1.1.3) Data validation Mechanical carriers include cylinders, coarse (1.2) Visual inspection groove gramophone, and instantaneous and vinyl discs. Table 1 summarizes the typologies of these carriers [25–30]. (1.3) Chemical analysis The main causes of deterioration are related to the insta- (1.4) Optimization of the carrier bility of mechanical carriers and can be summarized as fol- lows [25–27, 29]. (2) Signal transfer (1) Humidity. Humidity, as with all other data carriers, is the (2.1) Analysis of the /parameters most dangerous factor. While shellac and vinyl discs are less (2.2) System setup prone to hydrolytic instability, most kinds of instantaneous (2.2.1) Replay equipment (e.g., reel-to-reel tape discs are extremely endangered by hydrolysis. Additionally, recorder) all mechanical carriers may be affected by fungus growth, (2.2.2) Remediation equipment (converter, acqui- which occurs at humidity levels above 65%RH. sition software, monitoring, etc.) (2) Temperature. Elevated temperatures beyond 40 degrees (2.3) Monitoring Celsius are dangerous, especially for vinyl discs and wax (2.4) Data validation cylinders. Otherwise, the temperature determines the speed (2.5) Archival of the source carrier of chemical reactions like hydrolysis and should therefore be kept reasonably low and, most importantly, stable to avoid (3) Data processing and archival unnecessary dimensional changes.

(3.1) Metadata extraction (3) Mechanical Deformation. Mechanical integrity is of the (3.2) Completion of the preservative copy greatest importance for this kind of carriers. It is imperative that scratches and other deformation caused by careless Theinputoftheremediationprocessisanaudiodocu- operation of replay equipment are avoided. The groove that ment, and the expected output is its preservation copy—along carries the recorded information must be kept in an undis- with the source document ready to be stored again. After the torted condition. While shellac discs are very fragile, instan- remediation process, the carrier’s condition should normally taneous and vinyl discs are more likely to be bent by improper be better than before, thanks to the restoration to optimize its storage. Generally, all mechanical discs should be shelved 6 Journal of Electrical and Computer Engineering

Before playback Playback After playback

Original Preparation of Signal transfer Processing and Preservation carrier the carrier archiving copy

Control procedures

Figure 1: Scheme of the remediation process, in which three distinct steps can be observed, as well as the set of control procedures applied during the workflow. Each step is articulated in procedures and sub-procedures.

Table 1: Typologies of analogue mechanical carriers.

Carrier Period Composition Stocks Cylinder recordable 1886–1950s Wax 300,000 Wax and Nitrocellulose with plaster Cylinder replicated 1902–1929 1,500,000 (blue amberol) Mineral powders bound by organic binder Coarse groove disc replicated 1887–1960 10,000,000 (shellac) Coarse and microgroove discs recordable 1930–1950s Acetate or nitrate cellulose coating on aluminum 3,000,000 (“instantaneous discs”) (or glass, steel, card) Microgroove disc (vinyl) replicated 1948-today Polyvinyl chloride-polyacetate copolymer 30,000,000 vertically. The only exceptions are some soft variants of in- Table 2: Types of magnetic tape carriers. stantaneous discs. Type of Period Composition (4) Dust and Dirt. Dust and dirt of all kinds will deviate the recording pick-up stylus from its proper path causing audible cracks, Base: cellulose acetate clicks, and broadband noise. Fingerprints are an ideal adhe- 1935–1960 Analogue Magnetic pigment: Fe2O3 siveforforeignmatter.Adust-freeenvironmentandcleanli- Formats: open reel ness are, therefore, essential. For some examples on the effects Base: PVC on audio signals, see Figures 2 and 3. 1944–1960 Analogue Magnetic pigment: Fe2O3 2.6.2. Magnetic Tapes. The basic principles for recording sig- Formats: open reel nals on a magnetic medium were set out in a paper by Oberlin Base: polyester

Smith in 1880. The idea was not taken any further until Valde- 1959-today Analogue Magnetic pigment: Fe2O3 mar Poulsen developed his system in 1898. Formats: open reel, compact cassette IEC Magnetic tape was developed in Germany in the mid-1930s I to record and store sounds. The use of tape for sound record- Base: polyester ing did not become widespread, however, until the 1950s. 1969-today Analogue Magnetic tape can be either reel to reel or in cassettes. Table 2 /digital Magnetic pigment: CrO2 summarizes some types of these carriers. Formats: compact cassette IEC II, DCC The main causes of deterioration are related to the insta- Base: polyester bility of magnetic tape carriers and can be summarized as Magnetic pigment: SLH Ferro (blue tabs, 1976–1980 Analogue follows [25, 27, 29, 31–33]. Type I) and FeCr (red tabs, Type II) Formats: Elcaset (1) Humidity. Humidity is the most dangerous environmental factor. Water is the agent of the main chemical deterioration Base: polyester 1979-today Analogue process of polymers: hydrolysis. Additionally, high humidity /digital Magnetic pigment: metal particle values (above 65% RH) encourage fungus growth, which lit- Formats: compact cassette IEC IV, R-DAT erally eats up the pigment layer of magnetic tapes and floppy disks and also disturbs, if not prevents, proper reading of information. Floppy disks have been among the most com- usually memorized some short sound objects on floppy mon carriers for audio data storage in the field of electronic disks, synthesized at low sampling frequency (8–15 kHz). The musicinthe1980sand1990softhelastcentury.Composers study of these musical excerpts is very important from a Journal of Electrical and Computer Engineering 7

13.252 13.254 13.256 13.258 13.26 13.262 13.264 13.266 13.268 13.27 13.272 13.274 13.276 13.27813.28 13.282 13.284 1 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 −0.1 −0.2 −0.3 −0.4 −0.5 −0.6 −0.7 −0.8 −1

Figure 2: The figure shows an example of impulsive noise (click) in the waveform of an audio signal extracted from a 78 rpm shellac disc.It is clearly visible at second 13.263. Such disturbance can be causedbyaspeckofdustinadiscgroove.Time(seconds)isshownonthe𝑥-axis; amplitude (normalized) is shown on the 𝑦-axis.

17 17.1 17.2 17.3 17.4 17.5 17.6 17.7 17.8 17.9 18 18.1 18.2 18.3 18.4 18.5 18.6 18.7 18.8 18.9 19 19.1 19.2 19.3 19.4 19.5 19.6 19.7

12 k 10 k 9 k 8 k 7 k 6 k 5 k 4 k 3 k 2 k 0 k

Figure 3: The figure shows an example of broadband noise and impulsive noise in a spectrogram of an audio signal extracted froma78rpm shellac disc. Broadband noise is due to the granular nature of the material and its degradation over time. Time (seconds) is shown on the 𝑥- axis; frequency (Hertz) is shown on the 𝑦-axis. The shades of grey color show the amplitude: broadband noise is visible throughout the excerpt from the grey background (in contrast with the darker harmonic frequencies clearly visible up to ∼5 kHz); impulsive noises correspond to the vertical lines at seconds 18.08, 18.53, and 18.63.

musicological point of view. For instance, the multimedia but much needed in the preservation field. A number of anal- archive of the Centro di Sonologia Computazionale (CSC) yses are currently under processing in order to gain control of the University of Padova (http://csc.dei.unipd.it/)stores of parameters such as temperature and relative humidity in hundreds of floppy disks: it is unquestionably an outstanding relation with optimized storage environment conditions and testimony of the musical history in the 1980s and 1990s. carriers degradation. The analyses comprise Fourier Trans- form InfraRed (FTIR) Spectroscopic analysis in ATR, Ther- (2) Temperature. Temperature, coupled withrelative humidity, moGravimetric Analysis (TGA), Scanning Electronic Micro- is responsible for dimensional changes of carriers, which is scopy (SEM), Acetone extraction test, Acidity test, and X-Ray a particular problem for high-density tape formats. Temper- Diffraction. ature also determines the speed of chemical processes: the higher the temperature, the faster the chemical reaction (e.g., (3)MechanicalIntegrity.Mechanical integrity is a much hydrolysis). It is also important to consider the direct effect of underrated factor in the accessibility of data recorded on temperature and relative humidity on the magnetic particles, magnetic media: even slight deformations may cause severe but this behavior cannot be generalized because it depends on deficiencies in the playback process. Most careful handling (i) relative humidity and (ii) the type of iron oxide employed has to be exercised, along with regular professional mainte- by tape manufacturers, which is often not documented. In nance of replay equipment, which, in case of malfunctioning, 2011, the authors started a collaboration with the Department can destroy delicate carriers such as R-DAT very quickly. of Industrial Engineering—chemical sector of the University With all tape formats, it is most important to obtain an abso- of Padova—in order to exploit the knowledge and the tech- lutelyflatsurfaceofthetapepacktopreventdamagetothe niques of chemistry and materials science, which are scarce tape edges, which serve as mechanical references in the replay 8 Journal of Electrical and Computer Engineering of many high-density formats. All forms of tape should be Table 3: Recommended climatic storage parameters for mechanical stored upright. and for magnetic tapes. ± ± ± ± (4) Dust and Dirt. Dustanddirtpreventtheintimatecontact Temperature /24 h /year RH /24 h /year ∘ ∘ ∘ ∘ of replay heads to the medium, which is essential for the Preservation 5 C < t < 10 C ±1 C ±2 C30%±5% ±5% correct access to the information especially with high-density storage ∘ ∘ ∘ carriers. The higher the data density, the more cleanliness Access storage About 20 C ±1 C ±2 C40%±5% ±5 has to be observed. Even particles of cigarette smoke are big enough to hide information on modern magnetic formats. Table 4: Types of optical discs [35]. Also pollution caused by industrial smog can accelerate Type of chemical deterioration. The effective prevention of dust is Carrier Period Composition an indispensable measure for the proper preservation of recording magnetic media. Base: polycarbonate CD 1981-today Digital Reflective layer: (5) Magnetic Stray Fields. Magnetic stray fields are the natural aluminium, varnish, inks enemy of magnetically recorded information. Sources of dan- Base: polycarbonate gerous fields include dynamic microphones, loudspeakers, CD- 1992-today Digital Reflective layer: gold, silver, recordable andheadsets.Alsothesimplemagnetsusedformagnetic varnish, inks notice boards possess magnetic fields of dangerous magni- Base: polycarbonate tudes. By their nature, analogue audio recordings, including CD- 1996-today Digital Reflective layer: varnish, audiotracksonvideotapes,arethemostsensitivetomagnetic rewritable stray fields. inks Among the others, some effects can be Base: polycarbonate Reflective layer: silver SACD 1999-today Digital (i) “drop out” (i.e., the magnetic material fall off the tape); Coating: lacquer Base: polycarbonate (ii) “print-through” (i.e., a condition where low-frequen- cy signals on one tape winding imprint themselves on Inner layer: aluminium 2000-today the immediately adjacent tape windings); DVDA Digital Outer semireflective layer: gold (iii) “stretch” (i.e., the actual permanent stretching of the Coating: lacquer polyester caused by too tightly spooling the tape with noticeable pitch dropping). other environmental problems, described in Appendix C of Table 3 shows the correct parameters for the passive pre- [36]andin[37–39]. servation of mechanical carriers and magnetic tapes [26, 27, 29, 34]. (1) Discoloration (Includes Hazing, Bronzing, and Oxidation). A type of corrosion that affects the reflective layer, manifest- ing as a change in color on the surface. It may be caused by 2.6.3. Optical Discs. Optical disc recording technologies are aging or by temperature problems at the time the discs were based on the encoding of binary data in the form of pits and pressed. It usually starts at the edge of the disc and slowly lands on one of the disc’s surfaces. When read, pits and lands works its way towards the center. reflect the laser light projected along the data path: pits cor- respond to the binary value of 0, due to lack of reflection, and (2) Mold. “Mold usually takes the form of white or grey lands correspond to the binary value of 1, due to a reflection. patches on the surface, with a characteristic (fuzzy) structure The beam of light is emitted by a laser diode, usually inan visible under low-power magnification”36 [ ]. The onset of optical disc drive which spins the disc at speeds of about 200 mold is related to the humidity rate in the storage environ- to 4000 rpm or more, depending on the drive type, on the disc ment and the presence of other carriers affected by mold in format, and on the distance of the read head from the center thesameenvironment. of the disc. The most common optical audio carriers include Compact Discs (CDs, with a diameter of 12 centimeters), (3) Mechanical Integrity. Mechanical integrity is a fundamen- Mini CDs (with a diameter of 8 centimeters), and LaserDiscs tal factor in the accessibility of the data recorded on digital (LDs), besides the variants Super Audio CD (SACD) and optical media. Unlike in mechanical carriers and, to some Digital Versatile Discs Audio (DVDA) (see Table 4). extent, in magnetic tapes, where localized damage does not Digital information is pressed into a polycarbonate base, prevent access to other parts of the carrier, the loss of a set coated with a light reflective layer usually made of aluminum; of bits prevents that the rest of them are correctly read. Plus, however, gold and silver are also used. A transparent lacquer due to the functioning of the error correction algorithms, it is is placed over the reflective surface in order to protect it. Most hard to predict when the error rate will trespass the capacity opticaldiscsdonothaveanintegratedprotectivecasing,and of the algorithms to compensate such errors, and the carrier therefore they are easily subject to scratches, fingerprints, and will be suddenly irrecuperable. Mechanical integrity involves Journal of Electrical and Computer Engineering 9 damage ranging from scratches of various sizes and depths to Table 5: Recommended climatic storage parameters for optical cracks (breaks without physical separation), chips (missing media. small pieces, usually from the edge of the disc), and breaks Temperature ±/24 h ±/year RH ±/24 h ±/year (the disc has broken into distinct parts). ∘ ∘ ∘ Table 5 shows the recommended climatic access storage Preservation About 20 C ±1 C ±3 C40%±5% ±5% storage parameters for optical discs [35]. The behavior of digital optical media in their physical decay appears to reflect their “binary” nature: a disc will be Table 6: Types of [35]. either playable or unplayable, with very little difference in Type of Carrier Period Composition between (compared to the corruptions that can affect mag- recording netic tapes and that result in many levels of seriousness). Base: polycarbonate When a disc is corrupted but still playable, the extracted MD 1992–2013 Digital Reflective layer: signal will be often characterized by impulsive noises, some- replicated aluminium, varnish times intermitted with entire sets of consecutive missing sam- ples. Since monitoring during the process of signal extraction Base: polycarbonate MD Magneto-optical layer: is not provided for this type of carriers (see Section 3.2), 1992–2013 Digital it is mandatory to validate the data (either automatically or recordable ferromagnetic material manually) before storing it in a preservation copy, as the Reflective layer: aluminum extraction might terminate successfully but the extracted signal might nevertheless be corrupted. 3.1. Criteria for Selection. Before the remediation process starts, it is necessary to define a preservation schedule. This 2.6.4. Magneto-Optical Carriers. announced the Mini- depends on the number of documents, on their state of Disc (MD) audio format to appear in 1992, promising a com- preservation, on the archive priorities/policies, and on other bination of CD clarity with cassette convenience. Despite the factors. The preservation schedule is usually defined after the expectations, MiniDiscs did not do very well on the market study of the characteristics and of the state of preservation of andwereeventuallyovershadowedbysolid-statememory the carriers since these elements weigh in the decision. Deter- audio recorders. Altogether, the format survived two decades: mining a satisfactory schedule is not straightforward and SonyannouncedthatthedevelopmentofMDdeviceswillbe always requires a compromise, as the criteria to consider are halted in March 2013. often in contrast. It is revealed that the criteria suggested by The data is memorized in two steps: a laser heats oneof themaininstitutionsinthefieldhaveadifferentpriority: the sides of the disc, making the material in the disc suscep- IFLA (International Federation of Library Associations tible to a magnetic field, and on the other side, a magnetic and Institutions) [22] head alters the polarity of the heated area. Then the data (1) content: intellectual value of materials, their his- is accessed with the laser alone: taking advantage of the torical, scientific, and cultural significance (unique magneto-optic Kerr effect (MOKE), the player senses the sources must have priority); polarizationofthereflectedlightandisabletoreadthebinary values. (2) demand: priority is given to materials in constant Table 6 reports some manufacturing details about prere- demand; corded (replicated) and recordable MiniDiscs [35]. (3) condition: fragile and damaged unique materials (res- With regards to degradation, MiniDiscs are mainly toration procedures may be needed before the trans- threatened by dust and dirt deposits especially within the fer); housing, and in general they share the same problems of optical discs. IASA (International Association of Sound and Audiovi- sual Archives) [40] (1) documents in immediate risk/recordings on endan- 3. Methodology gered media; (2) documents that are part of an obsolete or commer- This section provides more details on the operational pro- cially unsupported system; tocol introduced in Section 2.5.Eachstepoftheprotocolis representedbyaflowchart,eachblockofwhichrepresents (3) documents in regular demand. an atomic task, eventually expanded in a separate flowchart. It is clear that the definition of the criteria for selection is Descriptionsandcommentsareasexhaustiveaspossible, an arbitrary choice and a mandatory one, the responsibility in order to minimize aleatory choices, and exceptions are of which belongs to the stakeholders, that is, the archival managed. In the next paragraphs a sample flowchart is institutions. Generally, the authors’ preference is a trade-off shown. between the IFLA and the IASA positions: Theprotocolhasbeenassessedandrefinedduringa number of research projects involving some of the finest (1) documents in immediate risk/recordings on endan- audioarchivesinItaly(seeSection 5). gered media; 10 JournalofElectricalandComputerEngineering

(2) demand: priority is given to materials in constant Source carrier demand; (3) documents that are part of an obsolete or commer- cially unsupported system. Y Clean with Dust? Once the preservation schedule is planned, the remedia- soft cloth tion process can start. N 3.2. Preparation of the Carrier. Thefirstthingtodowhen an audio document enters the remediation process is to document its physical condition, in order to know what was Y Provide its state before any restoration was performed (step (1.1)). The Reel ok? adequate reel photographical documentation (1.1.1) includes the carrier, its housing, the cover, the case/box, and any accompanying material. At choice, some of these may be acquired by means N of a scanner to enhance intelligibility (1.1.2). The photograph- ical documentation needs to be validated by the operator Take note of other sings ((1.1.3), in order to discard hazy or dark pictures) before mov- of physical degradation ing to the next step, aimed at detecting major sign of degra- dation of the physical carrier (step (1.2)), such as dirt/dust deposits, mold, and tear/breaks. Figure 4 shows the flowchart associated to this procedure. Leader Y Add leader tape Each block is extensively commented, and different com- ok? ments are provided for each carrier type. When possible, visual documentation is provided, such as pictures and short N demo movies. As an example, here are the instructions of the block “clean with soft cloth” for the carrier type Compact The carrier is ready for restoration Cassette. Figure 4: Flowchart of the step 1.2 preparation of the carrier → (i) Step 1. If dust is visible on the housing of the cassette, clean visual inspection. the housing before taking care of the tape. When the housing is free from dust, rewind the cassette.

(ii) Step 2. Flip the cassette over to whichever side is com- pletelyrewound:allofthetapeshouldbecoiledontheleft can be inferred from a direct analysis of the carrier and from coil, and the right coil should be empty. the writings on the cover and on the carrier itself, although often imprecise or missing. The methodology proposed in (iii) Step 3. Insert a pencil into the left spool, and then place this paper provides that when the noise reduction system is a soft microfiber cleaning cloth on top of the tape. Twist the unknown or it is not clear (even after a signal analysis and/or penciltotherightwhileslightlypressingdownonthetape a perceptual test) whether it has been used during the record- with the cloth. Stop after 10 to 30 turns and see if dust deposits ing, the carrier is played “flat,” without any compensation, arevisibleonthetapeoronthecloth.Ifso,keeptwistingthe andthischoiceisreportedinthepreservationcopydocu- penciltotherightwhileslightlypressingdownonthetape mentation. The analysis usually requires that the carrier is with the cloth until all of the tape has coiled up in the right tested on the reading device, which causes this operation to spool. Move the cloth every 30 to 50 turns in order to keep be particularly delicate. In the tower of Babel of the recording the tape in contact with a clean surface. Conversely, if there is formats, defining the correct ones is not an easy task. Some no dust on the tape, it does not need to be rewound by hand. historical research on the technologies used at the time of the The preparatory step terminates with the optimization recording may be required. A secondary aim of this test is to of the physical carrier (step (1.4)), achieved through specific detect some symptoms/corruptions that can only be detected restorative actions, in order to maximize its performance con- when the carrier is played (e.g., sticky shed syndrome for dition. The purpose of the optimization is to enable the ex- magnetic tapes). These symptoms still regard the physical traction of the best signal possible. This is of vital importance carrierandshouldbetreatedaccordinglybeforeproceeding for carriers in a poor state of preservation that might not with the signal extraction. endure multiple playback sessions. The definition of the recording format may have involved multiple replay devices: when the format is clear, the best 3.3. Signal Transfer. At this point, the carrier is physically equipment should be (1) selected and (2) adjusted for signal ready for playback. However, playback requires some addi- extraction (2.2.1). The same applies to the analog/digital con- tional information in order to read the carrier correctly. The verterandtherestoftheremediationchaindowntothework- recording format is specific for each carrier type (step (2.1)): it station that acquires the data (2.2.2). Some differences may Journal of Electrical and Computer Engineering 11 be observed in the procedure depending on the type of car- extracted with software tools: the best way to go is to extract rier, due to the mechanical-chemical-electrical specificities. them all with the assistance of a software tool, that will also The methodology presented in this paper aims at being gen- reorganize them in the predefined format, for example, XML eral; however detailing the procedure for each type of carrier (see Section 4). Finally, in order to complete the preservation goes beyond the scope of this work. At a macroscopic level, copy some more data processing is required (step (3.2)): a partition can be done between carriers with analog or digital images and schemes should be named and located correctly, signals. The main differences regarding these two groups are and the descriptive sheet should be compiled. At choice, a that carriers with digital signals do not require a link of the watermark could be added to images. These operations can be remediation chain that is fundamental for carriers with performed manually or with the assistance of software tools analog signals, the analog/digital converter; that some types (see Section 4). of carriers with digital signals are the only exception to moni- toring (e.g., CD, audio files, etc.); that algorithms for auto- matic error detection/correction can be used during the digi- 3.5. Preservation Copy. This subsection provides detailed tal-to-digital copy. information on the aforementioned preservation copy, which An important feature of the methodology proposed by is a key element in the methodology developed by the authors. the authors is that automatic rerecordings with simultaneous A preservation copy (or archive copy) is “the artifact desig- use of several systems is impossible, because the protocol natedtobestoredandmaintainedasthepreservationmaster. requires that each re-recording is monitored by an operator [omissis] Such a designation means that the item is used only (step (2.3)). Every audio document is inherently unstable and under exceptional circumstances” [41]. Audio carriers, espe- requires the annotation and the description of a number of cially modern high-density formats, are vulnerable by their signal alterations; let alone the supervision of the remediation very nature. In addition, there is always the risk of accidental chain (a malfunctioning device can tear, break, or crumple damage through improper handling, malfunctioning equip- thecarrier;areasonwhyanoperatormustalwaysbereadyto ment, or disaster. A possible strategy is the creation of “access intervene). Here is a list of the alterations that can be noted copies,” low-quality copies of preservation copies that can during playback: serveasanadjuncttothecatalogue,tohelpresearchersdecide what documents they wish to study. A copy of average/good (i) local noise: clicks, pops, signal dropout due to joints, quality may be acceptable for access in situ of the original. or tape degradation; Relying on copies (online or locally) to reduce the frequency (ii) global noise: hums, background noise, or distortion of access to the physical original will reduce the stress on the (periodical or nonperiodical); original and help in its preservation. A clear policy about the classes of researchers that are allowed access to physical ori- (iii) alterations produced when the sound was being ginals—particularly fragile ones—will also help in the doc- recorded: electrical noises (clicks, ripples), micro- uments survival. It is clearly impossible to totally restrict the phone distortions, blows on the microphone, or access to originals, but most users can carry out their research induction noise; using good quality access copies [25]. (iv) signal degradation due to malfunctions of the record- Restoration, when referred to a preservation copy, is only ing system (i.e., partial tracks deletion). allowed if it is intended to optimize the physical condition of the carrier before signal extraction. In the authors’ view, only When the signal extraction is complete, a set of manual the intentional alterations should be compensated at a pre- and of automatic controls should be applied to the digital servation copy level (e.g., correct equalization of the re-re- waveform (step (2.4)). If the resulting audio file is well cording system and decoding of any possible intentional sig- formed and compliant with the desired parameters, it can be nal processing interventions). As has been said in Section 2.2, exported in the format selected for the preservation action Schuller¨ suggests that also unintentional alterations are com- (see Section 3.5.1). At this point, all information has been pensated at a preservation copy level. Differing from this extracted from the original carrier (from the photographic position, the authors believe that unintentional alterations documentation to the audio signal), and it can exit the should not be compensated, since they witness the true history remediation system (step (2.5)). It should never be dismissed of the transmission of the audio document, representing the so- before validating the integrity of the preservation copy. As called “secondary information” introduced in Section 2.3.An has been said in Section 2.5,discardingasourcedocument example is bias frequency (the addition of an inaudible high- should never be undertaken lightly. Only the legal owner of frequency signal to the audio signal. Bias increases the signal the document can take the responsibility of such decision. In quality of audio recordings pushing the signal into the linear general, all the originals exit the remediation system in the zone of the tape’s transfer function [42]orbroadbandimpul- best condition possible and ready for long-term storage (e.g., sive noise (approximately ranging from 100Hz to 100kHz, it vacuum packing is desirable for magnetic tapes). can be caused by electrostatic discharges, by imperfections of the electrical/electronic circuits or by other damage of 3.4. Data Processing and Archival. Audio metadata are of electrical-chemical-mechanical nature). Both fall outside the paramount importance in the documentation of the preser- frequency range of the primary information (desired signal), vation copy (step (3.1)). Some of them are known apriori, andthisiswhythere-recordingmustbecarriedoutatthe11 some could be extracted manually, and some can only be highest standards available. 12 JournalofElectricalandComputerEngineering

During the remediation process, every part of the physical original document—multimedia in itself, because it consists Audio signal of audio, images (label, case, carrier corruptions, etc.), text 1st level: checksums (MD5, SHA1, CRC32) Metadata (accompanying material), and smell (mould, vinegar odor, 2nd level: file format specifications etc.)—is converted into a digital file, which results in a Uni- Pictures, transmediation Contextual information media document: a fusion of different media in a single flow system technical scheme of bits [43]. What cannot be directly represented in a digital form (i.e., smell) is thoroughly documented in the descriptive Descriptive sheet sheet. Figure 5 shows the logical structure of a preservation copy:itincludes(a)adescriptivesheetthatlistsallofthe files in the preservation copy, the provenance of the docu- Figure 5: Logical representation of the elements contained in a ment, the details about each audio file, and the venue of the preservation copy. transfer together with the person responsible for the creation of the copy; (b) the audio signal; (c) first-level metadata: checksum of the audio files; second-level metadata: technical 3.5.2. Video Shooting and Photographic Documents. The specifications of the file formats included in the preservation information reported on edition containers, labels, and other copy(bwf,pdf,etc.);(d)photographicaldocumentationof attachments should be stored with the preservation copy as a the carrier, its case, and the accompanying material and a static image, as well as clearly visible carrier corruptions (two technical sheet describing the transfer system. The purpose of examples are given in Figure 6). providing a preservation copy with this documentation meets A video of the carrier during playback—synchronized the requirement expressed in [44], that all compensations and with the audio signal—ensures the preservation of eventual processing, if applied, are “based on the capacity for precise information on the carrier, especially open-reel tapes (phys- counteraction” (which means reversibility of each operation ical conditions, presence of intentional alterations, corrup- and, consequently, capacity to trace the original characteris- tions, graphical signs). The video recording offers the follow- tics/values that were modified). ing. In a preservation copy, a distinction is made between (1) Information related to magnetic tape assembly oper- metadata and contextual information. Metadata indicates the ations and corruptions of the carrier (disc, cylinder, content-dependent information that can be automatically or tape), which are indispensable to distinguish the extracted from the audio signal; contextual information is the intentional from the unintentional alterations during additional content-independent information, such as a pho- the restoration process [8, 50, 51]. tographical documentationof the carrier case and the accom- panyingmaterial[45]. Two levels of metadata are also found (2) A description of the irregularities in the playback (see Figure 5): the first one is represented by metadata that speed of analogue recordings (such as wow and flutter, were extracted directly from the audio signal contained in the which are audio distortions perceived as an undesired preservation copy; the second one is represented by the doc- frequency modulation in the range of [52]: (i) wow umentation of the file formats contained in the preservation from 0.5 Hz to 6 Hz, (ii) flutter from 6 Hz to 100 Hz. copy (audio signal and contextual information, i.e., text, still The distortions are introduced to a signal by an irreg- images, etc.). ularvelocityoftheanaloguemedium.Astheirreg- ularities can originate from various mechanisms, the resulting parasitic frequency modulations can range 3.5.1. Format for the Audio Files. The audio signal should be from periodic to accidental, having different instan- stored in the preservation copy using the Broadcast Wave taneous values): in discs, and so on, a spindle hole Format, sampled at least at 96 kHz with a 24-bit resolution not precisely centered and/or the warping of the disc (for digital source documents, such as Compact Discs and cause a pitch variation; in tape recorders, an irreg- Digital Audio Tapes, the sampling frequency and resolution ular tape motion during playback (a change in the of the preservation copy can equal the original). It is advisable angularvelocityofthecapstanordraggingofthe to use the monophonic format, where each recording track tape within an audio cassette housing) causes changes is equivalent to a different file with Pulse Code Modulation in frequency. From the video, it is possible to locate representation [46, 47]. For further details on Broadcast Wave automatically the imperfections that occurred during Format refer to [48, 49]. the A/D transfer [45] and to distinguish them from These guidelines follow the precept that “the worse the the alterations already present in the recording, thus signal, the higher the resolution”,attributed to George Brock- increasing the information about the signal, useful for Nannestad during a personal communication to the authors restoration. in October 2007. The statement by Brock-Nannestad is based (3) Instructions for the performance of the piece (in on the fact that the characterization of the documents with particular in the electroacoustic music for tape): from a low-quality useful signal usually relies on its corruptions, thevideoanalysis,someprintsofthetapecanbe which are generally spread on a broad band. Therefore a high displayed; they represent either the synchronization resolution is needed in order to capture as much information of the score or the indication of particular sound about the corruptions. events. Journal of Electrical and Computer Engineering 13

(2) description of the preservation copy (general info and audio metadata for each audio file); (3) description of the source document (provenance, recording format, etc.); (4) description of the video recording, if present.

3.5.4. Data Integrity Verification. It is widely accepted that, for both technical and economic reasons, the preservation of audio documents relies upon transfer to, and storage in, the digital domain. However, digital files and carriers are not immune from format obsolescence and physical degradation. (a) Quite the opposite, the pace at which technology advances makes each new “generation” shorter, and a minor physical corruptioncanmakealltheinformationinaccessible(unlike analogue carriers, where a scratch on the surface of a phono- graphicdiscdoesnotpreventtheaccesstootherpartsofit).In this sense, digital files and carriers show weaknesses that are even more dangerous than the analogue ones, because “digital information can be lost—without warning—at any time” [54]. Such evidence calls for safety strategies. The most straightforward is having multiple copies of the documents, so that if one is corrupted or erased, another is available. The slogan “one copy is no copy” expresses this principle: a mini- mum of two copies must be available at all stages of the trans- (b) fer and archiving process [54]. Fortunately, the act of copying digital files does not incur the loss of fidelity inherent in Figure 6: (a) Example of one of the set of images included in the pre- servation copy of a compact cassette. Each element should be clearly analogue copying, so two or twenty copies do not deteriorate visible in all its details, so usually a preservation copy contains from the files quality. However, a corrupted file might be copied four to eight images. It is recommended that at least one includes several times, and the error(s) would not show until the file is the carrier with all its accompanying material—especially if these being read, which is too late. A possible solution is to extract arenotlistedsomewhereelseinthetextualdocumentation.Below andtosavethedocuments’checksumandtoverifythem the document’s signature, a small ruler can be observed: according periodically. This ensures data integrity during the preser- to the recommendation of the Italian Ministry of Cultural Heritage vation workflow or during storage or transmission. Our [53], this simple object conveys the real proportions of the depicted methodology provides that three different types of checksum elements. (b) Serious corruption of an optical disc: this image should of the audio files are extracted (MD5, SHA-1, and CRC32). be stored in the discs’ preservation copy in order to add information The terms “checksum” and “message digest” are commonly about the artifacts of the audio signal. used interchangeably. However, the term “checksum” is more correctly used for the product of a cyclical redundancy check (CRC32), whereas the term “message digest” refers to the The video file should be stored with the preservation copy. result of a cryptographic hash function (MD5, SHA-1). The selected resolution and the compression factor must at According to the preservation best practices expressed in leastallowtolocatethesignsandcorruptionsofthecarrier.In [1], the methodology proposed by the authors provides that the authors’ experience, a 320×240 pixel resolution video with the checksum values are stored within the preservation copy medium-quality DivX compression met this requirement. (in the descriptive sheet and in an XML file) as well as in the database, with backup copies kept in multiple physical locations. This avoids susceptibility to a single point of failure 3.5.3. Descriptive Sheet. Apreservationcopyismeanttobe in the system or other disasters. self-explicated. It should provide all the information needed to access, read, and interpret correctly its content in twenty or a hundred years from now. Apart from the technical speci- 4. Original Software Tools fications of the file formats, this aim is achieved with the inclusion of a descriptive sheet, which proves useful also in The software system presented in this section consists ofa case the logical structure of the archive is modified and some set of tools that support the remediation process. It is open documents get misplaced. A descriptive sheet is divided in source (GNU GPL v.3) and it includes two Java applications four sections: with a graphic user interface (GUI) and a number of shell scripts and Java programs without GUI. All of these elements (1) a complete list of the elements included and their are integrated in the preservation workflow schematized in relative path; Figure 7.Theuserstheyaredestinedtoaremembersofthe 14 JournalofElectricalandComputerEngineering preservation and the cataloguing staff (i.e., trained people involved in the preservation process). In function of the role played in the workflow, the soft- ware tools described in the next subsections can be grouped Staff for the description of the in the following: Preservation staff recordings content (1) tools for the active preservation of audio documents; (2) tools for the description of the contents (cataloguing); (3) tools for data monitoring and maintenance; Active preservation Cataloguing (4) tools for data sharing. The block on the top left in Figure 7,reading“activepre- servation”,includes all the steps of the protocol introduced in Section 2.5 and detailed in Section 3. It starts with the original Preservation Manual processing Manual Access archive document entering the remediation chain and it terminates archive material the sound of with the preservation copy transferred to the long-term stor- agesystem.Thearchiveofpreservationcopies,representedby the block on the bottom left in Figure 7, is the expected out- put. The software tools involved are described in Section 4.1. Control and backup procedures At this point, preservation could be considered complete, because the original documents have been safeguarded and Figure 7: High-level schematization of the preservation workflow the related data and metadata have been organized and from the viewpoint of the software tools. stored. But most archives express their mission as to “identify, acquire,andpreservearchivalmaterial[ongroundsoftheir enduring cultural, historical, or evidentiary value], and to described in Section 4.2 together with other automatized make it available [to the highest standards].” Therefore, pre- tasks mainly related to backups and data monitoring. servation should be intended in a broader sense that is not Regardless of the type of the recordings, the relation limited to storage: unrestricted access must be made available between the original document, the preservation copy, and “forever”—decades or centuries, or long enough to be con- the access copy is 1 : 1. This is not true for the relation between cerned about the obsolescence of technology [55]. the preservation/access copy and the audio resources for pub- As a consequence, additional steps are needed in order lic fruition, which are abstract entities independent from the to obtain an archive of catalogued audio resources ready to structure of the physical originals. They consist of audio files be accessed by the general public, starting from the archive of varying duration, corresponding to self-concluded acous- of preservation copies. The main difference between the two ticeventssuchasaninterview,asong,andanintermezzo. stages is the description of the content, which is missing Figure 8 showsthissituation:whiletheoriginaldocuments in the preservation copies. The preservation copies are only (represented by compact cassettes) and the preservation intended to safeguard the audio document as such, without copies (represented by the folders) coincide, the fruition units any relation to its content (symphonies, interview, electro- (represented by the yellow rectangles) may have an arbitrary acoustic music, or even silence). But the final users need to relation with the source documents. This is true for all search the documents with keywords related to the content archives, but it is more evident where the recordings have (title, author, subject, etc.). For this task, solid competences been gathered on the field, such as in a speech archive of on the content of the recordings are required; in this sense, linguistics: every inch of tape was used; therefore the record- this represents a variable part of the preservation process, ings are often split on multiple sides, and a side can contain becauseunlikethefirstpart,whichappliestoalltypes several recordings. The cataloguing staff listens to the digi- archives, this one depends on the area of interest of the tized audio and reorganizes it by cutting and merging tracks recordings: it may be formed by musicologists, anthropolo- to obtain a new set of audio files. This step is described in gists, linguists, historians, and so on. In this paper, a scenario Section 4.3. The expected output is an archive of resources with linguists is considered, that is, a sound archive of speech for fruition, represented by the block on the bottom right in documents. Going back to the workflow, the cataloguing staff Figure 7. Another important step of the preservation process need to access the audio contained in the preservation copies, is complete: the next—and last—is to design and build a (web) which is cumbersome and difficult to share over a network access system onto the archive of resources for fruition, to let connection (a stereo audio file with a duration of 48 minutes, final users perform searches, retrieve relevant information, a sampling frequency of 96 kHz, and a resolution of 24 bit and listen to the audio. Also the preservation metadata can occupies ∼1.5 GB). To solve this problem, access copies are be made available, but the audio in the preservation copies is created (see Section 3.5). These are made available to the cat- notintendedforcirculationandhasnorelationwhatsoever aloguing staff, which in our scenario is geographically distant with the access system. from the laboratory and need to download the files from It is to be noted that almost every step of the process can the internet in order to process them on their local work- be assisted or automatized by software tools, except for the stations.Thesestepsarecompletelyautomatizedandare reorganization of the audio material, represented by the pink Journal of Electrical and Computer Engineering 15

(1) the creation and the maintenance of the archive of preservation copies; (2) data ingestion into the database (see Section 4.4). The usefulness of PSKit PreservationPanel is mainly repre- sented by the quality control that it performs on the remedia- tion process, managing data and metadata in parallel and thus ··· ensuring a constant alignment between them. It significantly reduces the processing timing, by batch processing some categories of files, and at the same time eases the workload of 1 2 𝑁 the operator, by hiding all information that is known apriori or derivable. There are redundant controls associated with U1 U2··· UN everystepoftheworkflowcarriedoutbythesoftware.The interface of PSKit PreservationPanel is organized in panels, ascanbeseeninFigure 11. The panel for the description of single documents (physi- Figure 8: The relation between the audio documents (in the middle) cal original, recording format, and preservation copy) is cus- and the preservation copies (on top) is always 1 : 1. The relation between the latter and the fruition units (U1, U2, etc.) can be tomized for each carrier type. Filters are applied to attributes, complex, which is often the case with archives of recordings gathered that may apply or not to the selected carrier type, resulting on the field for linguistics and ethnomusicological studies. in different components on the interface. The introduction of errors is minimized by loading in the components only the valid values for each applicable attribute. For audio-specific metadata extraction (panel for batch processing), the authors rectangle in the middle of Figure 7, which must completely be have implemented in their system a modular tool for analysis carried out by human operators. Finally, as shown in Figure 7, and validation of digital objects in (digital) preservation a set of control procedures is always active during the entire programs developed by JSTOR and Harvard University [56]. process. The complete set of audio-specific metadata automatically From the viewpoint of their function, the software tools extractedwithJHoveislistedinTable 7. canbedividedtothefollowing. Other operations supported by PSKit PreservationPanel (i) Working tools in the panel for batch processing are (a) assignment of audio and contextual information files from default temporary (1) Archive alignment (working local archive, folders to the correct preservation copies; (b) creation of an mid-/long-term archive on the remote server, XML file with the checksums of the audio files for each pre- backups) servation copy; (c) creation of descriptive sheet; (d) valida- (2) Creation and sharing of access copies tion; and (e) transfer to remote server. Figure 9 schematizes (3) Database the actions that take place in the laboratory during the process of active preservation and that are assisted by Preservation- (4) Programs for data ingestion into the database Panel. (ii) Control tools 4.2. Automatic Processing. This subection describes the pro- (1) Process monitoring cedures that take place on the server machine to which the (2) Data verification (mid-/long-term) complete preservation copies have been transferred using PSKit PreservationPanel. All of the procedures are auto- (3) Backup procedures (database, archives) matically executed on a daily-weekly-monthly-yearly basis (4) Monitoring of the data growth according to the crontab job scheduler configuration:

4.1. PSKit PreservationPanel. Preservation Software Kit (1) Daily schedule (PSKit) PreservationPanel is an open-source software appli- (i) Creation of access copies cation developed in Java by the authors and licensed with the GNU GPL v.3. It counts almost 50,000 lines of code (ii) Sharing of access copies on a web page with andithasbeenusedsince2011inthelaboratoryforaudio limited access preservation of the Scuola Normale Superiore di Pisa (see (iii) Automatic mail messages to notify the new Section 5.2), where it assisted the creation of 217 preservation available audio files copies (analysis, organization, transfer, and archival), equal (iv) Calculation of the total duration of the digitized to 430 hours of digitized audio and to 700 GB, in just three audio working days—while the digitization of the audio required (v) Backup of the audio archive six months. (vi) Backup of the database At a user level, the main functions of PSKit Preservation- Panel are (2) Weekly schedule 16 JournalofElectricalandComputerEngineering

∗ Table 7: Metadata associated to the audio files in the descriptive sheet of a preservation copy. Items marked with astar( ) are not obtained with JHove: the name of the file is given by the user, and the duration of the file is calculated with PSKit PreservationPanel (class MetadatiAudio, private method durataTraccia(File f) that returns a String with the duration in the format HH:MM:SS).

Metadata Description ∗ Name Name of the audio file contained in the preservation copy including the extension, and without any leading directory components. ∗ Duration Duration of the audio file in the format HH:MM:SS. Size Size of the audio file in a human readable format, for example 874 MB or 1.2 GB. Extension Extension of the audio file, for example, wav. Format (MIME type) The standard recognized name for the format of the audio object. Encoding The encoding scheme used when audio digitization occurred for the described audio object. The majority of digital audio recordings will have a value of “PCM”. Profile String obtained with the combination of the values Format and Audio Format extracted from the header of the audio file. For example, PCMWAVEFORMAT. See the WAVE file format specifications for details. Tracks are here intended as channels; therefore a monorecording will have 1 track. This definition has been adopted to resolve the ambiguity of terminology between “track” (here intended as autonomous audio file, e.g., a CD has 14 tracks) and “channel” (typical values would be 1, 2, 4, and 8). Note that the metadata “Signal type” Number of tracks (mono, stereo,...) is not listed in this table because it is a qualitative evaluation of the operator and it cannot be extracted automatically—nevertheless it is included in the descriptive sheet in the section that describes the physical original. In the authors’ experience, “number of tracks,” “number of channels,” and “signal type” are the minimum combination of metadata to describe any type of recording format with no ambiguity whatsoever. The number of bits per sample for the audio content of the described audio object. This element describes the Bitdepth actual number of bits of the sample, whereas Word Size describes the number of bytes used to contain the sample. Sample rate The sample rate of the audio data for the described audio object. Byte order The order in which a sequence of bytes are stored in computer memory. Used to indicate whether the file isin little-endian or big-endian order. First sample offset The byte offset of the start of the data chunk which actually contains the waveform data with respect tothe beginning of the file. Compression Name of the algorithm for data compression applied to the audio file, if any. For preservation copies, it should always be “none.” Checksum MD5 A string indicating the checksum signature of the audio object. For example, 8847e21949079bfd4bf0c2bc26ba074a Checksum CRC32 A string indicating the checksum signature of the audio object. For example, 422f2901 Checksum SHA-1 A string indicating the checksum signature of the audio object. For example, 10a000e21a4c99479b15b852fcb3467b51c08cf5

(i) Backup of the website (ii) Possibility to monitor single processes with (ii) Grouping of daily database backups in compres- periodical reports sent via mail sed archives The creation of access copies (item number (1) in the pre- (3) Monthly schedule vious list) involves a shell script that downsamples the audio in the preservation copies and converts it to a compressed for- (i) Monitoring the new items added to controlled mat, adding new metadata to the header. The compressed files vocabularies in the database are then associated to the photographic documentation, and (ii) Grouping of weekly database backups in com- the resulting object is moved to the archive of access copies. pressed archives This archive is accessible through a web site with restricted access (item number (1) in the previous list). Finally, every (4) Yearly schedule morningascriptchecksifonthepreviousdaytherehavebeen (i) Grouping of monthly database backups in com- new uploaded preservation/access copies and, if so, sends pressed archives a notification mail message with the list of the audio files and link to retrieve them. This way, each member of the (5) Other services cataloguing staff is updated on a daily basis and can access the new documents very quickly, enabling a fast processing (i) Periodical messages reporting the status of the chain from active preservation to the archive of preservation server machine copies and to the archive for access. Journal of Electrical and Computer Engineering 17

4.4. Database. Adatabasethatisdesignedwithinthescope Laboratory of a preservation project must be able to maintain the data, as well as the relation between the data, that belong to two Signal extraction different and opposite aspects of the process: the active pre- servationofaudiodocumentsontheonehandandthecata- Contextual loguing of the contents on the other. These aspects are oppo- information Metadata site because the first focuses on the document as such, regard- lessoftheircontent.Themetadataproducedinthispartofthe

Long-term process are mainly technical and audio specific. The second archive Optimization Processing and aspect sacrifices the fidelity to the data structure of the physi- (carrier restoration) data validation caloriginalcarrierinfavorofabstractself-concludedacoustic events that coincide with culturally mediated interpretations oftheaudiostream.Themetadataproducedinthispartofthe Documents ready Documents for treatment after treatment process are content dependent and vary according to the area of interest of the recordings. For example, in the scenario with Storage area speech documents of dialectological relevance, the metadata will include the linguistic area, the date of creation and Figure 9: Abstract view of the activities that take place in the thepeopleinvolvedintheconversation,thetopicsofthe preservation laboratory. conversation, and an abstract of what is being said, in order to provide the final users with as much information as possible. The database adopted by the software system described in this paper was created using the relational model, with Oracle The system as it has been described in this work uses a MySQL. Its population is performed by means of the applica- redundant Hard Disk Drive array to store the data (archive of tions described in Sections 4.1 and 4.3, which are able to reach preservation copies, access copies, database, etc.) in the long the database from a local or a remote network connection. term. The specific problems related to the checking, The relation between the preservation copy and the audio files refreshing, and migration technology fall within the scope for fruition is maintained by associating the identification of the research area of digital preservation (see, e.g., number of the source audio files (which contain a reference [57]). tothepreservationcopytheybelongto)totheidentification number of the audio files for fruition. Additional information 4.3. Cataloguing. The tool that the members of the cata- regards the starting/ending time of the reorganized audio, loguing staff use to populate the database is called PSKit which allows the recreation of the fruition file in case it gets CataloguingPanel and like PSKit PreservationPanel has been lost or erased. More generally, the metadata stored in the developed in Java during the collaboration with the sound database allows the reconstruction of every operation per- archive of the Scuola Normale Superiore di Pisa (see formed starting from the preservation copies, which are the Section 5.2). ones that should always be safeguarded with the highest care. Once the audio material has been reorganized and new Thanks to the database design, the final users will be able audio files coinciding with self-concluded acoustic events to retrieve the technical audio-specific data starting from a have been created, the cataloguing staff can proceed with its search by content, and vice versa, meeting the needs of the classification and its description. PSKit CataloguingPanel is users interested in the recordings content, those interested essentially an interface for data ingestion, but just like PSKit in preservation, and the more general public searching the PreservationPanel it follows an attentive study of the require- archive out of curiosity and for personal entertainment. ments carried out in tight collaboration with the linguistics research team. Long discussions have been made to refine the data model, trying to bridge the gab between two disciplines 5. Case Studies such as computer science and linguistics: an extended time spent for the collection of the requirements has paid off with The operational protocol presented in this paper has been the realization of a tool that guides the operator in the work- developed in several research projects to which the authors flow making it fast and simple and at the same time ensures have participated since 1996: “Electronic Storage and Preser- data consistency and minimizes the introduction of errors. vation of Artistic and Documentary Audio Heritage (speech PSKit CataloguingPanel allows to maintain the connec- and music)” funded by the National Research Council of Italy tion between the new self-concluded audio files coinciding (CNR); “Preservation and Online Access of Contemporary withthecataloguedacousticevents,andthesourcefilesinthe Music Italian Archive” funded by the Italian Ministry for preservation copies that have been used to created them. This Scientific Research; “Preservation and Online Fruition of the connection is crucial because it is the only link between the Audio Documents from the European Archives of Ethnic archive for preservation and the archive for access, that is, Music” funded by the EU under the Program Culture2000; between the preservation metadata and the content descrip- “Search in Audio-Visual Content Using Peer-to-Peer Infor- tions. More details about this relation are provided in the next mation Retrieval” funded by the EU under the Sixth Frame- section. work Programme. 18 JournalofElectricalandComputerEngineering

(2) definition of an operational protocol for the remedia- tion of the audio documents and for the management of the laboratory (maintenance routines for technical equipment, rules for the documents handling and storage during treatment, etc.); (3) realization of a laboratory, inside the archive, fully equipped to support the active preservation of audio documents (Figure 10); (4) knowledge transfer to the archive personnel (on a methodological level, and on an applied level: use of the technical equipment, use of the original software tools developed on purpose during the project, phys- Figure 10: Laboratory for the preservation and the restoration of audio documents inside the audiovisual archive of the Fondazione ical restoration of the audio carriers, etc.); Arena di Verona. On the right, a large collection of recording/replay (5) creation of over 1,200 preservation copies of different devices can be observed on the shelves. On the left, an open-reel types of audio documents (magnetic tapes, optical recorder Studer A812. On the desk, an A/D-D/A Converter PRISM discs, and digital nonaudio carriers). ADA-8XR and two Compact Cassette recorders. The archive comprises tens of thousands of audio docu- ments stored on different carriers (from wax cylinders to dig- ital carriers), nearly a hundred pieces of equipment for replay and for recording (from wire to magnetic tape recorders and ) and bibliographic publications (including monographs and all issues of more than sixty music journals from the 1940s to 1999). The audio archive is divided in a historical section, containing the live recordings of the operas staged every year at the Arena during the summer season, and the Mario Vicentini section, named after its donor and theestimatedvalueofwhichis2,300,000Euros.Thearchive is being enriched every year with the recordings of the new opera seasons, now memorized on digital data storage devices. Figure 11: Panel for the description of single documents in PSKit Most recordings consist in live performances of classical PreservationPanel for the carrier type “phonographic disc.” operas staged in an open-air setting, the Arena di Verona. On the average, the audio documents presented a good state of preservation, except for some among the oldest open-reel tapes(late1960sandearly1970s),alotofCompactCassettes Important European archival institutions have been from 1981 to 1983, and Compact Discs from the early 2000s involved, including “Speech and Music Archives” of the which proved unreadable. The precision incubator com- National Research Council of Italy “Archive of the Studio di monly used for the thermic treatment of tapes with specific Fonologia Musicale,”owned by the Italian National Broadcast syndromes [32, 58, 59]hasbeenlargelyused,andwithsatis- Television; “Luigi Nono Archive”; “Bruno Maderna Archive;” factory results in nearly all cases. The large number of docu- and “Historical Archive of Contemporary Arts” of the Venice ments stimulated the development of original software tools Biennial. to control and to automatize various aspects of the preser- In particular, the protocol has been perfected during two vation process and most importantly to perform periodical research projects presented in the next paragraphs. controls over the archive of preservation copies in order to avoid inconsistencies with very negative consequences on 5.1. 2009–2011: Arena di Verona. In 2009, the audiovisual the information retrieval system. Manual control is out of archive of the Fondazione Arena di Verona and the Depart- question because, by the end of the project, the preservation ment of Computer Science of the University of Verona started copies almost reached 2 TB, equal to over 18,000 single files. a collaboration finalized at the digitization of the sound col- Four pieces of original software have been developed by lection owned by Arena. The joint project lasted two years and theauthors,andallofthemarestillinuseatthearchive accomplished the following goals: over a year after the end of the project without need of further assistance. To learn more about the project, see [60] (1) definition of a methodology for preservation, after a for a description of the methodology and of the partial survey on the state of preservation of the archive and results obtained in the first year of the project and [61] of its peculiarities (number and type of documents, for a more detailed description of the software utilities genre of the recordings, objectives of the digitization, developed during the second year of the project and the final etc.); results. Journal of Electrical and Computer Engineering 19

5.2. 2011–2013: Scuola Normale Superiore di Pisa. Audio re- of linguistics holds its own collection of audio recordings, cordings play an important role in the field of linguistics: from but the research project had the ambition of censusing and life stories to dialect investigations, they provide researchers collectingasmanyexternalpublicandprivatearchivesaspos- with invaluable first hand material. Unfortunately first hand sible on the territories where the Tuscan language (vernacolo does not mean being unchanged with respect to the original toscano) is spoken. This results in a highly heterogeneous acoustic source: capturing an acoustic event is never a neutral set of documents, differing in number, recording format, operation, as explained in Section 1.Particularyforthose decade of creation, state of preservation, and quality of the areas of linguistics that employ frequency analysis (formant description of the content. The archives that accepted to col- and pitch detection), the requirements of accuracy, relia- laborate (over twenty to date, and more joining) sent their bility, and authenticity of the audio material, mentioned in material over to the laboratory, where it would be processed Section 2.3, cannot be renounced. Aware of the risks of audio and eventually returned to their original location. This is a material with uncertain origin, the research team of the good example of a scenario where the original document is Laboratory of Linguistics at the Scuola Normale Superiore di not under the responsibility of the preservation staff after the Pisa started a project in collaboration with the University of treatment, which shows the importance of creating preserva- Siena finalized at tion copies that contain absolutely complete information and of the best quality possible, since it is not possible to perform (1) applying the methodology for preservation to the extra controls, comparisons and analysis—just as in the sce- audio documents of the area of linguistics, after a narios where the physical original is lost, stolen, or destroyed. study of their characteristics (most problems are Tolearnmoreabouttheproject,see[62] for a description related to the fact that the recordings have been gath- of the methodology specifically adapted for an archive of ered on the field, often in inadequate conditions, with speech documents and [63] for a description of the software nonprofessional equipment and inexpensive carriers, system that has been presented in its final form in the present etc.); work. (2) realizing a restoration laboratory, inside the labora- tory of linguistics, fully equipped to support the active preservation of audio documents; 6. Conclusions (3) transferring knowledge to the technical staff of labo- From the viewpoint of the scientific research, the original ratory of linguistics to enable autonomy after the end software tools that constitute the system presented in this of the project (on a methodological level and on an paper belong to the area of cultural interfaces. They are applied level: use of the equipment, use of the original characterized by a development that has been conducted with softwaretoolsdevelopedonpurposeduringthepro- a transdisciplinary approach, achieved by means of a tight ject, physical restoration of the audio carriers, etc.); and prolonged collaboration of the researchers in computer (4) creating an archive of preservation copies of different science and engineering with the experts of different scientific types of audio documents (open-reel tapes, Compact disciplines (musicologists, linguists, archivists, etc.). This col- Cassettes, Digital Audio Tapes, and digital nonaudio laboration created the conditions for many opportunities of carriers). mutual confrontation and for a better modeling of the re- quirements, for a deeper understanding of the others’ termi- Even if the importance of audio documents is acknowl- nology, methodology, and so forth. This approach was mainly edged in the community of linguistics, this project has been reflected by the first ever on the Italian territory to introduce the compe- tences of preservation into a laboratory of linguistics. For over (1) the design of the database (reconciliation of different ayear,theauthorshaveworkedinsidethelaboratoryintight approaches to preservation, information modeling); connection with the linguistics research team, in a proac- tive multidisciplinary attitude, bridging the gap between (2) the formalization of the workflow (accordance be- the disciplines’ approaches and vocabularies. As a result, the tween the theory of preservation and laboratory prac- original software system described in Section 4 was devel- tice): sustainability and processing timing, assistance oped, with the subsequent benefits of reducing processing to remediation sessions, and cataloguing on parallel timing,ofenablingthetreatmentofagreaternumberofdoc- workstations. uments, and of gaining complete control over the data growth and integrity. The software system has been in use 24/7 since Besides improving the quality of the laboratory work, the September 2011, and it is currently being used by the preser- results obtained prove that the introduction of original sof- vation staff (2 people based in Pisa) and by the cataloguing tware tools in the process of active preservation of audio doc- staff (5 people based in other cities of Italy, exchanging data uments opens the way for an effective answer to the method- through a network connection). By the time this paper is ological questions of reliability with relation to the record- being written, a total of 759 preservation copies have been ings as documentary sources, also clarifying the concept of created, equal to 1.7 TB; 10,000 single files; and 945 hours of “faithfulness” to the original and situating it in the precise re-recorded audio. limitations of the audio equipment technology. Amajorquestionthathadtobesolvedisrelatedtoitem The architecture of the software system reflects the found- number (4) in the previous list of objectives: the laboratory ing principles of the methodology applied to the active 20 JournalofElectricalandComputerEngineering

preservation audio documents. Assuming that those princi- [16] Premis project (preservation metadata: Implementation strate- plesareagreeduponandrespected,thesystemcanbeadapted gies), http://www.loc.gov/standards/premis/. to the needs of other archives at low cost—a desirable objec- [17] Speech & language data repository (sldr), http://crdo.up tive that would help correct preservation practices to spread, .univ-aix.fr/. encouraged by the use of adequate and freely shared tools. [18] http://www.dei.unipd.it/bressanf/,2013. [19] W. Storm, “The establishment of international re-recording Conflict of Interests standards,” Phonographic Bulletin,vol.27,pp.5–12,1980. [20] G. Brock-Nannestad, “The objective basis for the production of high quality transfers from pre-1925 sound recordings,” in AES The authors have no conflict of interests to declare in relation ∘ with the software system described in this work. Preprint n 4610 Audio Engineering Society 103rd Convention,pp. 26–29, AES, New York, NY, USA, 1997. [21] D. Miller, The Science of Musical Sounds, Macmillan, New York, References NY, USA, 1922. [22] IFLA, “Audiovisual and multimedia section: guidelines for [1] M. Casey and B. Gordon, Sound Directions: Best Practices for digitization projects: for collections and holdings in the public Audio Preservation, Harvard University, Cambridge, Mass, domain, particularly those held by libraries and archives,”Tech. USA; Indiana University, Bloomington, Ind, USA, 2007. Rep., International Federation of Library Associations and [2] R. Edmonson, Memory of the World: General Guidelines to Safe- Institutions (IFLA), The Hague, The Netherlands, 2002. guard Documentary Heritage, UNESCO, Paris, France, 2002. [23] L. Duranti, “Interpares3—team Canada final report,”Tech. Rep., [3] T. W. Adorno, Philosophy of New Music, University of Min- University of British Columbia, 2012. nesota Press, Minneapolis, Minn, USA, 2006. [24] M. Factor, E. Henis, D. Naor et al., “Authenticity and provenance [4] E. Cohen, Preservation of Audio in Folk Heritage Collections in long term digital preservation: modeling and implemen- in Crisis, Council on Library and Information Resources, tation in preservation aware storage,” in Proceedings of the Washington, DC, USA, 2001. 1st Workshop on the Theory and Practice of Provenance,San [5] C. A. Paton, “Preservation re-recording of audio recordings in Francisco, Calif, USA, February 2009. archives: problems, priorities, technologies, and recommenda- [25] IFLA/UNESCO, “Safeguarding our documentary heritage/con- tions,” The American Archivist, vol. 61, no. 1, pp. 188–219, 1998. servation preventive du patrimoine documentaire/salvaguard- [6] A. M. Karafin, “Digitization of sound recordings as an example andonuestro patrimonio documental,” CD-ROM Bi-lingual: for preservation of oral and music folklore heritage,”in Proceed- English/French/Spanish. UNESCO “Memory of the World” ings of the 1st International Conference on The Future of Informa- Programme, French Ministry of Culture and Communication, tion Sciences (INFuture ’07): ‘Digital Information and Heritage’, 2000. pp.139–152,Zagreb,Croatia,2007. [26] S. Laurent, The Care of Cylinders and Discs,MiltonKeynes,1997. [7] A. Smith, Why Digitize? CouncilonLibraryandInformation [27] A. Pickett and M. Lemcoe, Preservation and Storage of Sound Resources, Washington, DC, USA, 1999. Recordings, ARSC, Washington, DC, USA, 1991. [8] AES, “AES recommended practice for audio preservation and [28] S. Khanna, “Vinyl compound for the phonographic industry,” restoration—storage and handling—storage of polyester-base Journal of Audio Engineering Society,vol.25,no.10-11,pp.712– magnetictape,”Tech.Rep.AES22-1997(r2008),AES,NewYork, 717, 1977. NY, USA, 1997. [9] AES, “AES standard for audio preservation and restoration— [29] M. Calas and J. Fountaine, La Conservation des Documents magnetic tape—care and handling practices for extended Sonores, CNRS, Paris, France, 1996. usage,”Tech. Rep. AES49-2005, AES, New York, NY,USA, 2005. [30] L. Burt, “Chemical technology in the edison recording indus- [10] AES, “AES Information document for preservation of audio try,” Journal of Audio Engineering Society,vol.25,no.10-11,pp. recordings—extended term storage environment for multiple 712–717, 1977. media archives,”Tech. Rep. AES-11id-2006, AES, New York, NY, [31] H. N. Bertram and E. F. Cuddihy, “Kinetics of the humid aging USA, 2006. of magnetic recording tape,” IEEE Transactions on Magnetics, [11] D. Schuller,¨ “The ethics of preservation, restoration, and re- vol. 27, pp. 4388–4395, 1982. issues of historical sound recordings,” JournalofAudioEngi- [32] G. Gibson, “Magnetic tape deterioration: recognition, recovery neering Society,vol.39,no.12,pp.1014–1016,1991. and prevention,” http://www.unesco.org/webworld/ramp/html/ [12] G. Boston, Safeguarding the Documentary Heritage. A Guide to r9704e/r9704e11.htm,1996. Standards, Recommended Practices and Reference Literature [33] G. Knight, “Factors relating to long term storage of magnetic Related to the Preservation of Documents of All Kinds, UNESCO, tape,” Phonographic Bulletin,no.18,pp.16–37,1977. Paris, France, 1988. [34] IFLA/UNESCO, “Safeguarding our Documentary Heritage/ [13] ZKM, “The digital oblivion: substance and ethics in the conser- Conservation preventive du patrimoine documentaire,” chap- vation of computer-based art,” in Proceedings of the Interna- ter: Magnetic carriers, UNESCO, Boston, Mass, USA, http:// tional Symposium of the Research Project Digital Art Conser- webworld.unesco.org/safeguarding/EN/pdf/txt opti.pdf, 2000. vation, Zentrum fur¨ Kunst und Medientechnologie Karlsruhe [35] IFLA/UNESCO, “Safeguarding our Documentary Heritage/ (ZKM), 2010. Conservation preventive du patrimoine documentaire,” chap- [14] Planet project (planets, preservation and long-term access ter: Optical carriers, UNESCO, Boston, Mass, USA, http:// through networked services), http://www.planets-project.eu/. webworld.unesco.org/safeguarding/EN/pdf/txt opti.pdf, 2000. [15] Presto space project (preservation towards storage and access. [36] IASA, The IASA Cataloguing Rules, IASA Editorial Group, 1999. standardised practices for audiovisual contents in europe), [37] M. Sueiro, AVDb—Audio and Moving Image Survey Tool, http://www.prestospace.org/. Columbia University Libraries, New York, NY, USA, 2008. Journal of Electrical and Computer Engineering 21

[38] M. Casey, “Facet (field audio collection evaluation tool)— evaluating potential strategies and building preservation plans,” format characteristics and preservation problems version 1.0,” International Journal on Digital Libraries,vol.10,no.4,pp.133– Tech. Rep., Indiana University, Bloomington, Ind, USA, 2008, 157, 2010. http://www.dlib.indiana.edu/projects/sounddirections/facet/ [58] E. F. Cuddihy, “Aging of magnetic recording tape,” IEEE Trans- facet formats large.pdf. actions on Magnetics, vol. 16, no. 4, pp. 558–568, 1980. [39] N. Film, “Technical glossary,” of Australia, S. A., http://nfsa [59] J. Bigourdan, J. M. Reilly, K. Santoro, and G. Salesin, “The pre- .gov.au/preservation/glossary/,2013. servation of magnetic tape collections: a perspective,” Tech. [40] IASA Technical Committee, “The safeguarding of the audio Rep. NEH GRANT PA-50123-03, Image Permanence Institute, heritage: ethics, principles and preservation strategy,”Tech. Rep. Rochester Institute of Technology, Rochester, NY, USA, 2006. IASA-TC 03, IASA Technical Committee, 2005. [60] F. Bressan, A. Roda, S. Canazza, and F. Fontana, “Toward an [41] IASA Technical Committee, “Guidelines on the production and informed procedural approach to the preservation of audio do- preservation of digital objects,” Tech. Rep. IASA-TC 04, IASA cuments: the case of the “fondazione arena di verona” archive,” Technical Committee, 2004. in Proceedings of Sharing Cultures 2011—2nd International Con- [42] 3M Company, “High frequency bias requirements for magnetic ference on Intangible Heritage, S. Lira, R. Amoeda, and C. Pin- tape recording,” 3M SoundTalk Bulletins,vol.1,no.2,pp.1–4, heiro, Eds., pp. 177–185, Green Lines Institute, Tomar, Portugal, 1968. 2011. [43] N. Negroponte, Being Digital, Vintage Books, New York, NY, [61] F. Bressan and S. Canazza, “Towards a procedure for quality USA, 1995. control over large collections of digitized audio data: the case of [44] D. Schuller,¨ “Preserving the facts for the future: principles and the“FondazioneArenadiVerona”,”inMultimedia for Cultural practices for the transfer of analog audio documents into the Heritage. First International Workshop (MM4CH ’11),C.Grana digital domain,” Journal of the Audio Engineering Society,vol. andR.Cucchiara,Eds.,vol.247ofCommunications in Computer 49,no.7-8,pp.618–621,2001. and Information Science, pp. 103–113, Springer, Berlin, Germany, [45] N. Orio, L. Snidaro, S. Canazza, and G. L. Foresti, “Methodolo- 2012. gies and tools for audio digital archives,” International Journal [62] S. Canazza, S. Calamai, P. M. Bertinetto, and A. de Dominicis, on Digital Libraries,vol.10,no.4,pp.201–220,2009. “A protocol for the preservation of speech document archives: [46] IASA Technical Committee, “The safeguarding of the audio toward the digital curation of the Carta dei dialetti italiani,” heritage: ethics, principles and preservation strategy,”Tech. Rep. in Proceedings of the 5th International Congress on ‘Science IASA-TC 03, IASA Technical Committee, 2005. and Technology for the Safeguard of Cultural Heritage in the Mediterranean Basin’,vol.4,pp.84–93,2012. [47] M. Hart, “Preserving our musical heritage: a musician’s out- reach to audio engineers,” Journal of the Audio Engineering [63] F. Bressan, “Un ambiente informatico per la gestione dei pro- Society,vol.49,no.7-8,pp.667–670,2001. cessi relativi alla conservazione attiva dei documenti sonori,” in Proceedings of the 19th International Colloquium on Music Infor- [48] AES, “AES standard on network and file transfer of audio— matics (Colloqui di Informatica Musicale, CIM) “Sonic Synesthe- audio-file transfer exchange—file format for transferring digital sia”,pp.34–39,November2012. audio data between systems of different type and manufacture,” Tech. Rep. AES31-2-2006, AES, New York, NY, USA, 2006. [49] EBU, “Specification of the Broadcast Wave Format: a format for audio data files in broadcasting,” Tech. Rep. 3285, EBU, Grand- Saconnex, Switzerland, 1997. [50] S. Canazza, Noise and Representation Systems: A Comparison among Audio Restoration Algorithms, Lulu Enterprise, Raleigh, NC, USA, 2007. [51] S. Canazza and A. Vidolin, “Preserving electroacoustic music,” Journal of New Music Research,vol.30,no.4,pp.351–363,2001. [52] A. E. Society, “Method for measurement of weighted peak flut- ter of sound recording and reproducing equipment,” AES StandardAES6-2008,AES,NewYork,NY,USA,2008. [53] R. Galasso and E. Giffi, “La documentazione fotografica delle schede di catalogo—metodologie e tecniche di ripresa,” Tech. Rep., Istituto Centrale per il Catalogo e la Documentazione (ICCD), Ministero per i Beni e le Attivita` Culturali, 1998. [54] D. Schuller,¨ “The ethics of preserving audio and video docu- ments,” Report 2004-2005, pp. 78–80, Information for All Pro- gramme (IFAP), 2006. [55] M. Hedstrom, “It’s about time: research challenges in digital archiving and long-term preservation. final report on a work- shop on research challenges in digital archiving,” Tech. Rep., Library of Congress, Washington, DC, USA, 2002. [56] Jhove, “Jstor/harvard object validation environment,” Tech. Rep., http://hul.harvard.edu/jhove/ . [57] C. Becker, H. Kulovits, M. Guttenbrunner, S. Strodl, A. Rauber, and H. Hofman, “Systematic planning for digital preservation: International Journal of

Rotating Machinery

International Journal of Journal of The Scientific Journal of Distributed Engineering World Journal Sensors Sensor Networks Hindawi Publishing Corporation Hindawi Publishing Corporation Hindawi Publishing Corporation Hindawi Publishing Corporation Hindawi Publishing Corporation http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014

Journal of Control Science and Engineering

Advances in Civil Engineering Hindawi Publishing Corporation Hindawi Publishing Corporation http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014

Submit your manuscripts at http://www.hindawi.com

Journal of Journal of Electrical and Computer Robotics Engineering Hindawi Publishing Corporation Hindawi Publishing Corporation http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014

VLSI Design Advances in OptoElectronics International Journal of Modelling & International Journal of Simulation Aerospace Navigation and in Engineering Observation Engineering

Hindawi Publishing Corporation Hindawi Publishing Corporation Hindawi Publishing Corporation Hindawi Publishing Corporation Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Hindawi Publishing Corporation http://www.hindawi.com Volume 2014

International Journal of International Journal of Antennas and Active and Passive Advances in Chemical Engineering Propagation Electronic Components Shock and Vibration Acoustics and Vibration

Hindawi Publishing Corporation Hindawi Publishing Corporation Hindawi Publishing Corporation Hindawi Publishing Corporation Hindawi Publishing Corporation http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014